這個 ComfyUI 工作流程可從單張圖像生成短片，並精確遵循使用者自訂的鏡頭或主體運動路徑。它採用 Wan 2.1 I2V ATI 模型，將文字提示、參考圖像與軌跡 JSON 結合，讓鏡頭移動或物件運動不再隨機。核心節點包含 UNETLoader（Wan2_1-I2V-ATI-14B_fp8_e4m3fn.safetensors）與 VAELoader（wan_2.1_vae.safetensors）作為擴散主幹，CLIPLoader 與 CLIPTextEncode 處理提示詞，CLIPVisionLoader 與 CLIPVisionEncode 處理圖像條件，WanTrackToVideo 則負責軌跡控制。KSampler（由 ModelSamplingSD3 驅動）進行去雜訊，VAEDecode 將潛變量轉為影格，CreateVideo 與 SaveVideo 組合並輸出影片。
技術上，流程會用 CLIPTextEncode 編碼正/負向提示詞，並用 CLIPVisionEncode 從輸入圖像提取視覺特徵。PrimitiveStringMultiline 節點存放從 Trajectory-Annotation-Tool 匯出的軌跡 JSON；WanTrackToVideo 讀取該 JSON，將每一影格的運動控制注入取樣過程。ModelSamplingSD3 設定排程器，讓 KSampler 以與 Wan 2.1 訓練相容的方式去雜訊。取樣後，VAEDecode 重建影格，CreateVideo 以指定解析度與 FPS 組成影片。以 4090D 24GB 顯卡在 720x480 下，首次運行約 500 秒，權重快取後約 463 秒（實際時間依設定而異）。
API
Use this workflow from code
Every Comfy workflow is a JSON graph. The payload below is this workflow, exactly as ComfyUI runs it — fetch it from the URL, keep it in version control, or load it in ComfyUI and run it node by node.
Run it from TypeScript or Python with the Comfy SDK. The same code targets Comfy Cloud or a ComfyUI you host yourself — only the base URL changes.
// Install (beta)
npm i @comfyorg/sdk
// Run "Wan ATI" (TypeScript)
import { Comfy } from "@comfyorg/sdk";
const client = new Comfy({ apiKey: "comfyui-..." });
// This workflow, exported in API format (see note below)
const wf = await client.workflows.fromFile("video_wan_ati_api.json");
const asset = client.assets.fromFile("input.png");
wf.setInput("240", "image", asset); // LoadImage
wf.setInput("6", "text", "your prompt here"); // CLIPTextEncode
const job = await client.run(wf);
await job.getOutputs("258")[0].toFile("output.png"); // SaveVideoThe SDK takes a workflow in API format: open this workflow in ComfyUI and use File → Export Workflow (API).
Comfy Cloud API access requires a plan with an API key.
常見問題
